The conventional approach to valuing agricultural research is the economic surplus approach (Alston et al 1995). This approach is commonly applied to country wide research programs rather than disaggregated to district level. In this note, we demonstrate how to use causal machine learning heterogeneous treatment effects to make return on investment analyses that are sufficiently disaggregated.
2 Steps in Return on Investment Economic Surplus Model
# library(ggplot2)# ggplot() +# geom_line(data = subset(APY_data, APY_data$Crop == "Wheat"), aes(x = Year, y = Area_Hectare, group = Districts, linetype = Districts, color = Districts), linewidth = 1.5) +# labs(x = "Year", y = "Area") +# theme_bw() # Having the baseline we need to create this from 2020 to 2030library(tidyverse)APY_data_wheat_2019_30 <- APY_data_wheat_2015_19_ave %>%mutate(year =list(2019:2030)) %>%separate_rows(year, sep =",", convert = T)# Delete Kishanganji due to lack of dataAPY_data_wheat_2019_30 <-subset(APY_data_wheat_2019_30, !(APY_data_wheat_2019_30$Districts =="KISHANGANJ"))
2.2 Step 2: Adoption data and projections
Code
# District level Cumulative Adoption -------------------------------------------library(rio)DistrictCumulativeAdoption=import("DistrictCumulativeAdoption.csv") # Manuallibrary(nlme)DistrictCumulativeAdoption100=nlsList(cumulative_percent~100/(1+exp(-(a + b*herbi_1st_use))),start=list(a =1, b =0),data=groupedData(cumulative_percent~herbi_1st_use|A.q103_district,DistrictCumulativeAdoption),na.action=na.omit)summary(DistrictCumulativeAdoption100)
# Subset the years 2019 to 2030 and three columns Herb_adoption_2019_30=subset(Herb_adoption,Herb_adoption$year%in%c(2019:2030))ggplot(Herb_adoption_2019_30,aes(y=Herb_adopt_perc,x=year,color=Districts)) +geom_line()
2.3 Step 3: Supply, demand elasticities, and prices
Code
APY_plus_adoption$Supply_Elast=0.22# Taken from Kumar et al 2016APY_plus_adoption$Demand_Elast=0.340# Taken from Kumar et al 2011Prices=import("tabpricedist.csv")Prices=subset(Prices,select=c("A.q103_district","mean"))names(Prices)[1:2]=c("Districts","Prices")Prices$Districts=toupper(Prices$Districts)APY_plus_adoption=merge(APY_plus_adoption,Prices,by="Districts")
2.4 Step 4: Yield and cost changes due to research
# Producer surplus#APY_plus_adoption$producer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*APY_plus_adoption$K_shift_A*(1-0.5*APY_plus_adoption$K_shift_A*APY_plus_adoption$Supply_Elast)APY_plus_adoption$producer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*(APY_plus_adoption$K_shift_A-APY_plus_adoption$Z_t)*(1+0.5*APY_plus_adoption$Z_t*APY_plus_adoption$Demand_Elast)APY_plus_adoption$producer_surplus_Rs_per_ha=(APY_plus_adoption$producer_surplus/APY_plus_adoption$Area_Hectare)*1000# Consumer surplusAPY_plus_adoption$consumer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*APY_plus_adoption$Z_t*(1+0.5*APY_plus_adoption$Z_t*APY_plus_adoption$Demand_Elast)# Total surplusAPY_plus_adoption$total_surplus=APY_plus_adoption$producer_surplus+APY_plus_adoption$consumer_surplus
2.6.1 Small open economy producer surplus
For a small open economy in which the state’s contribution cannot affect global prices, producer surplus can be computed as: \[
\Delta PS_t = P_0Q_0K_t(1-0.5K_t \epsilon)
\]
India_aoi_sf$Districts <-toupper(India_aoi_sf$NAME_2)APY_plus_adoption_indic_sf <-merge(India_aoi_sf, APY_plus_adoption_indic, by ="Districts")# Average wheat area# Using tmaplibrary(tmap)tmap_mode("plot")Wheat_area_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Wheat_area_mean", palette ="viridis", scale =2.5, title ="Wheat area (ha)")Wheat_area_sf_tmap
Code
tmap_save(Wheat_area_sf_tmap, "figures/Wheat_area_sf_tmap.png", dpi =300, width =4.88, height =3.16)# Average wheat yieldstmap_mode("plot")Wheat_yield_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Yield_Tonnes_Hectare_mean", palette ="viridis", scale =2.5, title ="Wheat yields (t/ha)")Wheat_yield_sf_tmap
Code
tmap_save(Wheat_yield_sf_tmap, "figures/Wheat_yield_sf_tmap.png", dpi =300, width =4.88, height =3.16)# Average wheat productiontmap_mode("plot")Wheat_production_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Production_Tonnes_mean", palette ="viridis", scale =2.5, title ="Wheat production (t)")Wheat_production_sf_tmap
# Using tmaplibrary(tmap)tmap_mode("plot")prod_surplus_changes_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_net_present_value_sum_annual", palette ="viridis", scale =2.5, title ="Producer surplus (INR 1000)")prod_surplus_changes_sf_tmap
Code
tmap_save(prod_surplus_changes_sf_tmap, "figures/prod_surplus_changes_sf_tmap.png", dpi =300, width =3.88, height =3.16)# Small open economylibrary(tmap)tmap_mode("plot")prod_surplus_changes_open_econ_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_present_value_open_econ_sum_annual", palette ="viridis", scale =2.5, title ="Small open economy PS (INR 1000)")prod_surplus_changes_open_econ_sf_tmap
Code
tmap_save(prod_surplus_changes_open_econ_sf_tmap, "figures/prod_surplus_changes_open_econ_sf_tmap.png", dpi =300, width =3.88, height =3.16)# Producer surplus per ha# Using tmaplibrary(tmap)tmap_mode("plot")prod_surplus_changes_perha_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_net_present_value_sum_annual_Rs_ha", palette ="viridis", scale =2.5, title ="Producer surplus per ha")prod_surplus_changes_perha_sf_tmap
Code
tmap_save(prod_surplus_changes_perha_sf_tmap, "figures/prod_surplus_changes_perha_sf_tmap.png", dpi =300, width =3.88, height =3.16)## Small open economylibrary(tmap)tmap_mode("plot")prod_surplus_changes_open_econ_perha_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_present_value_open_econ_sum_annual_Rs_ha", palette ="viridis", scale =2.5, title ="Small open economy PS per ha")prod_surplus_changes_open_econ_perha_sf_tmap
# Yield changesYield_changes <-subset(Yield_changes, !(Yield_changes$Districts =="KISHANGANJ"))library(mapview)Yield_changes_sf <-merge(India_aoi_sf, Yield_changes, by ="Districts")mapview(Yield_changes_sf, zcol ="Yield_gain_t_ha", layer.name ="Yield gain to herbicides (tons per ha)")
Code
# Using tmaplibrary(tmap)tmap_mode("plot")Yield_changes_sf_tmap <-tm_shape(Yield_changes_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Yield_gain_t_ha", palette ="viridis", scale =2.5, title ="Yield gain to herbicides (tons per ha)")Yield_changes_sf_tmap
The notebook has demonstrated the workflow for calculating spatially disaggregated economic surplus measures that can be used to target and prioritize agricultural interventions.
5 References
Alston, J.M., Norton, G.W., and Pardey, P.G. 1995. “Science under scarcity: Principles and practice for agricultural research evaluation and priority setting.” CAB International. Wallingford, UK.
Source Code
---title: "Returns on Investment to Herbicides: Example of Spatial Economic Surplus Approach"format: html: code-fold: true code-tools: truefig-dpi: 300fig-width: 8.88fig-align: centerfig-height: 5self-contained: trueauthor: Maxwell Mkondiwaeditor: visualtoc: truetoc-location: leftnumber-sections: trueexecute: message: false warning: false echo: true---# IntroductionThe conventional approach to valuing agricultural research is the economic surplus approach (Alston et al 1995). This approach is commonly applied to country wide research programs rather than disaggregated to district level. In this note, we demonstrate how to use causal machine learning heterogeneous treatment effects to make return on investment analyses that are sufficiently disaggregated.# Steps in Return on Investment Economic Surplus Model## Step 1: Baseline area, production, and yield```{r, warning=FALSE}library(rio)APY_data <- import("Data_1997_23_Cleaned.xlsx", sheet = "All_data")APY_data$Production_Tonnes <- as.numeric(APY_data$Production_Tonnes)APY_data_wheat_2015_19 <- subset(APY_data, APY_data$Season == "Rabi" & APY_data$Crop == "Wheat" & APY_data$Year %in% c("2015-16", "2016-17", "2017-18", "2018-19", "2019-20"))APY_data_wheat_2015_19$Season <- NULLAPY_data_wheat_2015_19$Crop <- NULLAPY_data_wheat_2015_19$Season <- NULLlibrary(data.table)APY_data_wheat_2015_19 <- data.table(APY_data_wheat_2015_19)APY_data_wheat_2015_19_ave <- APY_data_wheat_2015_19[, lapply(.SD, base::mean, na.rm = TRUE), by = .(Districts)]APY_data_wheat_2015_19_ave <- APY_data_wheat_2015_19[, .(Area_Hectare = base::mean(Area_Hectare), Production_Tonnes = base::mean(Production_Tonnes), Yield_Tonnes_Hectare = base::mean(Yield_Tonnes_Hectare)), by = .(Districts)]# Download all the datalibrary(reactable)library(htmltools)library(fontawesome)htmltools::browsable( tagList( tags$button( tagList(fontawesome::fa("download"), "Download as CSV"), onclick = "Reactable.downloadDataCSV('APY_data_wheat_2015_19_ave', 'APY_data_wheat_2015_19_ave.csv')" ), reactable( APY_data_wheat_2015_19_ave, searchable = TRUE, defaultPageSize = 38, elementId = "APY_data_wheat_2015_19_ave" ) ))# library(ggplot2)# ggplot() +# geom_line(data = subset(APY_data, APY_data$Crop == "Wheat"), aes(x = Year, y = Area_Hectare, group = Districts, linetype = Districts, color = Districts), linewidth = 1.5) +# labs(x = "Year", y = "Area") +# theme_bw() # Having the baseline we need to create this from 2020 to 2030library(tidyverse)APY_data_wheat_2019_30 <- APY_data_wheat_2015_19_ave %>% mutate(year = list(2019:2030)) %>% separate_rows(year, sep = ",", convert = T)# Delete Kishanganji due to lack of dataAPY_data_wheat_2019_30 <- subset(APY_data_wheat_2019_30, !(APY_data_wheat_2019_30$Districts == "KISHANGANJ"))```## Step 2: Adoption data and projections```{r, warning=FALSE}# District level Cumulative Adoption -------------------------------------------library(rio)DistrictCumulativeAdoption=import("DistrictCumulativeAdoption.csv") # Manuallibrary(nlme)DistrictCumulativeAdoption100=nlsList(cumulative_percent~100/(1 + exp(-(a + b*herbi_1st_use))),start=list(a = 1, b = 0), data=groupedData(cumulative_percent~herbi_1st_use|A.q103_district,DistrictCumulativeAdoption),na.action=na.omit)summary(DistrictCumulativeAdoption100)alphaDistrictCumulativeAdoption100 <- coef(DistrictCumulativeAdoption100) write.csv(alphaDistrictCumulativeAdoption100, file = "alphaDistrictCumulativeAdoption100.csv")# SSLOGISSSlogDistrictCumulativeAdoption100=nlsList(cumulative_percent~SSlogis(herbi_1st_use, 100, xmid, scale),start=list( xmid=1992, scale=5), data=groupedData(cumulative_percent~herbi_1st_use|A.q103_district,DistrictCumulativeAdoption),na.action=na.omit,pool=FALSE)summary(SSlogDistrictCumulativeAdoption100)nlssummary=summary(SSlogDistrictCumulativeAdoption100)xmid=nlssummary$coefficients[,,1]colnames(xmid) <- paste("xmid", colnames(xmid), sep = "_")scale=nlssummary$coefficients[,,2]colnames(scale) <- paste("xmid", colnames(scale), sep = "_")SSlogDistrictCumulativeAdoption100Results <-data.frame(coefficients(SSlogDistrictCumulativeAdoption100),xmid,scale)library(broom)library(tibble)library(dplyr)library(tidyr)library(purrr)SSlogDistrictCumulativeAdoption100Results$Slopes=1/SSlogDistrictCumulativeAdoption100Results$scaleSSlogDistrictCumulativeAdoption100Results$Intercept=-(SSlogDistrictCumulativeAdoption100Results$xmid/SSlogDistrictCumulativeAdoption100Results$scale)SSlogDistrictCumulativeAdoption100Results$Year10percent=((-2.2-SSlogDistrictCumulativeAdoption100Results$Intercept)/SSlogDistrictCumulativeAdoption100Results$Slopes)SSlogDistrictCumulativeAdoption100Results$Year90percent=((2.2-SSlogDistrictCumulativeAdoption100Results$Intercept)/SSlogDistrictCumulativeAdoption100Results$Slopes)SSlogDistrictCumulativeAdoption100Results$Years10_90=SSlogDistrictCumulativeAdoption100Results$Year90percent-SSlogDistrictCumulativeAdoption100Results$Year10percentwrite.csv(SSlogDistrictCumulativeAdoption100Results, file = "SSlogDistrictCumulativeAdoption100Results.csv")herbi_1st_use <- with(DistrictCumulativeAdoption, seq(2000,2030, length.out = 31))A.q103_district=levels(as.factor(DistrictCumulativeAdoption$A.q103_district))herbi_1st_use_dist=expand.grid(herbi_1st_use=herbi_1st_use,A.q103_district=A.q103_district)herbi_1st_use_dist$pred=predict(SSlogDistrictCumulativeAdoption100, herbi_1st_use_dist)herbi_1st_use_dist_Data=merge(DistrictCumulativeAdoption,herbi_1st_use_dist,by=c("A.q103_district","herbi_1st_use"),all.y=TRUE)write.csv(herbi_1st_use_dist_Data, file = "herbi_1st_use_dist_DataPredictionResults100.csv")Herb_adoption=subset(herbi_1st_use_dist_Data, select=c("A.q103_district","herbi_1st_use","pred") )names(Herb_adoption)[1:3]=c("Districts","year","Herb_adopt_perc")write.csv(Herb_adoption, file = "Herb_adoption.csv")# Plotting predicted herbicide adoption ratesggplot(Herb_adoption,aes(y=Herb_adopt_perc,x=year,color=Districts)) + geom_line()+ theme_bw()+ labs(y="Ädoption rate (% of farmers)")# Subset the years 2019 to 2030 and three columns Herb_adoption_2019_30=subset(Herb_adoption,Herb_adoption$year%in%c(2019:2030))ggplot(Herb_adoption_2019_30,aes(y=Herb_adopt_perc,x=year,color=Districts)) + geom_line()Herb_adoption_2019_30$Districts=toupper(Herb_adoption_2019_30$Districts)APY_plus_adoption=merge(Herb_adoption_2019_30,APY_data_wheat_2019_30,by=c("Districts","year"))APY_plus_adoption$Area_Hectare_Herb=(APY_plus_adoption$Herb_adopt_perc/100)*APY_plus_adoption$Area_Hectare```## Step 3: Supply, demand elasticities, and prices```{r, warning=FALSE}APY_plus_adoption$Supply_Elast=0.22 # Taken from Kumar et al 2016APY_plus_adoption$Demand_Elast=0.340 # Taken from Kumar et al 2011Prices=import("tabpricedist.csv")Prices=subset(Prices,select=c("A.q103_district","mean"))names(Prices)[1:2]=c("Districts","Prices")Prices$Districts=toupper(Prices$Districts)APY_plus_adoption=merge(APY_plus_adoption,Prices,by="Districts")```## Step 4: Yield and cost changes due to research```{r, warning=FALSE}library(rio)Yield_changes=import("tau.hat_weeding_herb_predictions_dist.csv")Yield_changes=subset(Yield_changes,select=c("A.q103_district","mean"))names(Yield_changes)[1:2]=c("Districts","Yield_gain_t_ha")Yield_changes$Districts=toupper(Yield_changes$Districts)APY_plus_adoption=merge(APY_plus_adoption,Yield_changes,by="Districts")# cost changesAPY_plus_adoption$Cost_changes=0```## Step 5: Research induced supply shift```{r, warning=FALSE}# K_shift_yld_comptAPY_plus_adoption$K_shift_yld_compt=APY_plus_adoption$Yield_gain_t_ha/APY_plus_adoption$Supply_Elast# K_shift_cost_componentAPY_plus_adoption$K_shift_cost_compt=APY_plus_adoption$Cost_changes/(1+(APY_plus_adoption$Yield_gain_t_ha/APY_plus_adoption$Yield_Tonnes_Hectare))# K_shiftAPY_plus_adoption$K_shift=APY_plus_adoption$K_shift_yld_compt-APY_plus_adoption$K_shift_cost_comptAPY_plus_adoption$K_shift_A=APY_plus_adoption$K_shift*APY_plus_adoption$Herb_adopt_perc*0.01# Z_t=K_t e/e+etaAPY_plus_adoption$Z_t=(APY_plus_adoption$K_shift_A*APY_plus_adoption$Supply_Elast)/(APY_plus_adoption$Supply_Elast+APY_plus_adoption$Demand_Elast)```## Step 6: Compute producer surplus, consumer surplus and total surplusFor a small closed economy, the equations producer surplus, consumer surplus and total surplus.Producer surplus$$\Delta PS_t = P_0Q_0 (K-Z)(1+0.5Z \eta)$$Consumer surplus $$\Delta CS_t = P_0Q_0 Z(1+0.5Z \eta)$$Total surplus $$\Delta TS_t =\Delta PS_t + \Delta CS_t $$```{r, warning=FALSE}# Producer surplus#APY_plus_adoption$producer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*APY_plus_adoption$K_shift_A*(1-0.5*APY_plus_adoption$K_shift_A*APY_plus_adoption$Supply_Elast)APY_plus_adoption$producer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*(APY_plus_adoption$K_shift_A-APY_plus_adoption$Z_t)*(1+0.5*APY_plus_adoption$Z_t*APY_plus_adoption$Demand_Elast)APY_plus_adoption$producer_surplus_Rs_per_ha=(APY_plus_adoption$producer_surplus/APY_plus_adoption$Area_Hectare)*1000# Consumer surplusAPY_plus_adoption$consumer_surplus=APY_plus_adoption$Production_Tonnes*APY_plus_adoption$Prices*APY_plus_adoption$Z_t*(1+0.5*APY_plus_adoption$Z_t*APY_plus_adoption$Demand_Elast)# Total surplusAPY_plus_adoption$total_surplus=APY_plus_adoption$producer_surplus+APY_plus_adoption$consumer_surplus```### Small open economy producer surplusFor a small open economy in which the state's contribution cannot affect global prices, producer surplus can be computed as: $$\Delta PS_t = P_0Q_0K_t(1-0.5K_t \epsilon)$$```{r}APY_plus_adoption$producer_surplus_open_econ <- APY_plus_adoption$Production_Tonnes * APY_plus_adoption$Prices * (1-0.5* APY_plus_adoption$K_shift_A * APY_plus_adoption$Supply_Elast)APY_plus_adoption$producer_surplus_open_econ_Rs_per_ha <- (APY_plus_adoption$producer_surplus_open_econ / APY_plus_adoption$Area_Hectare) *1000```## Step 7: Cost of research```{r, warning=FALSE}APY_plus_adoption$Cost_of_research=0```## Step 8: Compute economic indicators```{r, warning=FALSE}# Discount rateAPY_plus_adoption$discount_rate <- 0.2# timeAPY_plus_adoption$time <- APY_plus_adoption$year - 2019# cashflowsAPY_plus_adoption$net_producer_surplus <- APY_plus_adoption$producer_surplus - APY_plus_adoption$Cost_of_researchAPY_plus_adoption$net_producer_surplus_open_econ <- APY_plus_adoption$producer_surplus_open_econ - APY_plus_adoption$Cost_of_researchAPY_plus_adoption$net_consumer_surplus <- APY_plus_adoption$consumer_surplus - APY_plus_adoption$Cost_of_researchAPY_plus_adoption$net_total_surplus <- APY_plus_adoption$total_surplus - APY_plus_adoption$Cost_of_research# Calculate the Present Value of Each Cash FlowAPY_plus_adoption$Producer_present_value <- APY_plus_adoption$net_producer_surplus / ((1 + APY_plus_adoption$discount_rate)^APY_plus_adoption$time)APY_plus_adoption$Producer_present_value_open_econ <- APY_plus_adoption$net_producer_surplus_open_econ / ((1 + APY_plus_adoption$discount_rate)^APY_plus_adoption$time)APY_plus_adoption$Consumer_present_value <- APY_plus_adoption$net_consumer_surplus / ((1 + APY_plus_adoption$discount_rate)^APY_plus_adoption$time)APY_plus_adoption$Total_present_value <- APY_plus_adoption$net_total_surplus / ((1 + APY_plus_adoption$discount_rate)^APY_plus_adoption$time)# Sum all the Present Valueslibrary(data.table)APY_plus_adoption <- data.table(APY_plus_adoption)APY_plus_adoption_indic <- APY_plus_adoption[, .( Producer_net_present_value_sum = base::sum(Producer_present_value, na.rm = TRUE), Producer_present_value_open_econ_sum = base::sum(Producer_present_value_open_econ, na.rm = TRUE), Consumer_net_present_value_sum = base::sum(Consumer_present_value, na.rm = TRUE), Total_net_present_value_sum = base::sum(Total_present_value, na.rm = TRUE), Wheat_area_mean = base::mean(Area_Hectare, na.rm = TRUE), Area_Hectare_Herb_mean = base::mean(Area_Hectare_Herb, na.rm = TRUE), Production_Tonnes_mean = base::mean(Production_Tonnes, na.rm = TRUE), Yield_Tonnes_Hectare_mean = base::mean(Yield_Tonnes_Hectare, na.rm = TRUE)), by = .(Districts)]# Producer surplus small open economyAPY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual <- APY_plus_adoption_indic$Producer_present_value_open_econ_sum / 11sum(APY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual)summary(APY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual)APY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual_Rs_ha <- (APY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual / APY_plus_adoption_indic$Wheat_area_mean) * 1000summary(APY_plus_adoption_indic$Producer_present_value_open_econ_sum_annual_Rs_ha)# Producer surplusAPY_plus_adoption_indic$Producer_net_present_value_sum_annual <- APY_plus_adoption_indic$Producer_net_present_value_sum / 11sum(APY_plus_adoption_indic$Producer_net_present_value_sum_annual)summary(APY_plus_adoption_indic$Producer_net_present_value_sum_annual)APY_plus_adoption_indic$Producer_net_present_value_sum_annual_Rs_ha <- (APY_plus_adoption_indic$Producer_net_present_value_sum_annual / APY_plus_adoption_indic$Wheat_area_mean) * 1000summary(APY_plus_adoption_indic$Producer_net_present_value_sum_annual_Rs_ha)# Consumer surplusAPY_plus_adoption_indic$Consumer_net_present_value_sum_annual <- APY_plus_adoption_indic$Consumer_net_present_value_sum / 11sum(APY_plus_adoption_indic$Consumer_net_present_value_sum_annual)summary(APY_plus_adoption_indic$Consumer_net_present_value_sum_annual)APY_plus_adoption_indic$Consumer_net_present_value_sum_annual_Rs_ha <- (APY_plus_adoption_indic$Consumer_net_present_value_sum_annual / APY_plus_adoption_indic$Wheat_area_mean) * 1000summary(APY_plus_adoption_indic$Consumer_net_present_value_sum_annual_Rs_ha)# Total surplusAPY_plus_adoption_indic$Total_net_present_value_sum_annual <- APY_plus_adoption_indic$Total_net_present_value_sum / 11sum(APY_plus_adoption_indic$Total_net_present_value_sum_annual)summary(APY_plus_adoption_indic$Total_net_present_value_sum_annual)APY_plus_adoption_indic$Total_net_present_value_sum_annual_Rs_ha <- (APY_plus_adoption_indic$Total_net_present_value_sum_annual / APY_plus_adoption_indic$Wheat_area_mean) * 1000summary(APY_plus_adoption_indic$Total_net_present_value_sum_annual_Rs_ha)# Benefit-cost ratio# APY_plus_adoption$BCR=APY_plus_adoption$producer_surplus/APY_plus_adoption$Cost_of_research# Internal Rate of Return (IRR)# IRR(cf0,cf,times,plot=FALSE)# MIRR=# irr1 <- project1_cf %>%# select(cf) %>%# .[[1]] %>%# irr()# Modified Internal Rate of Return (MIRR)```# Spatial visualization of the economic surpluses```{r}library(geodata)India <-gadm(country ="IND", level =2, path ="shp")plot(India)India_aoi <-subset(India, India$NAME_1 =="Bihar")plot(India_aoi)plot(India_aoi, add =TRUE)library(sf)India_aoi_sf <-st_as_sf(India_aoi)library(mapview)mapview(India_aoi_sf)India_aoi_sf$Districts <-toupper(India_aoi_sf$NAME_2)APY_plus_adoption_indic_sf <-merge(India_aoi_sf, APY_plus_adoption_indic, by ="Districts")# Average wheat area# Using tmaplibrary(tmap)tmap_mode("plot")Wheat_area_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Wheat_area_mean", palette ="viridis", scale =2.5, title ="Wheat area (ha)")Wheat_area_sf_tmaptmap_save(Wheat_area_sf_tmap, "figures/Wheat_area_sf_tmap.png", dpi =300, width =4.88, height =3.16)# Average wheat yieldstmap_mode("plot")Wheat_yield_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Yield_Tonnes_Hectare_mean", palette ="viridis", scale =2.5, title ="Wheat yields (t/ha)")Wheat_yield_sf_tmaptmap_save(Wheat_yield_sf_tmap, "figures/Wheat_yield_sf_tmap.png", dpi =300, width =4.88, height =3.16)# Average wheat productiontmap_mode("plot")Wheat_production_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Production_Tonnes_mean", palette ="viridis", scale =2.5, title ="Wheat production (t)")Wheat_production_sf_tmaptmap_save(Wheat_production_sf_tmap, "figures/Wheat_production_sf_tmap.png", dpi =300, width =4.88, height =3.16)# Producer surplusmapview(APY_plus_adoption_indic_sf, zcol ="Producer_net_present_value_sum_annual", layer.name ="Producer surplus (INR 1000)")# Using tmaplibrary(tmap)tmap_mode("plot")prod_surplus_changes_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_net_present_value_sum_annual", palette ="viridis", scale =2.5, title ="Producer surplus (INR 1000)")prod_surplus_changes_sf_tmaptmap_save(prod_surplus_changes_sf_tmap, "figures/prod_surplus_changes_sf_tmap.png", dpi =300, width =3.88, height =3.16)# Small open economylibrary(tmap)tmap_mode("plot")prod_surplus_changes_open_econ_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_present_value_open_econ_sum_annual", palette ="viridis", scale =2.5, title ="Small open economy PS (INR 1000)")prod_surplus_changes_open_econ_sf_tmaptmap_save(prod_surplus_changes_open_econ_sf_tmap, "figures/prod_surplus_changes_open_econ_sf_tmap.png", dpi =300, width =3.88, height =3.16)# Producer surplus per ha# Using tmaplibrary(tmap)tmap_mode("plot")prod_surplus_changes_perha_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_net_present_value_sum_annual_Rs_ha", palette ="viridis", scale =2.5, title ="Producer surplus per ha")prod_surplus_changes_perha_sf_tmaptmap_save(prod_surplus_changes_perha_sf_tmap, "figures/prod_surplus_changes_perha_sf_tmap.png", dpi =300, width =3.88, height =3.16)## Small open economylibrary(tmap)tmap_mode("plot")prod_surplus_changes_open_econ_perha_sf_tmap <-tm_shape(APY_plus_adoption_indic_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Producer_present_value_open_econ_sum_annual_Rs_ha", palette ="viridis", scale =2.5, title ="Small open economy PS per ha")prod_surplus_changes_open_econ_perha_sf_tmaptmap_save(prod_surplus_changes_open_econ_perha_sf_tmap, "figures/prod_surplus_changes_open_econ_perha_sf_tmap.png", dpi =300, width =3.88, height =3.16)mapview(APY_plus_adoption_indic_sf, zcol ="Consumer_net_present_value_sum_annual", layer.name ="Consumer surplus (INR 1000)")mapview(APY_plus_adoption_indic_sf, zcol ="Total_net_present_value_sum_annual", layer.name ="Total economic surplus (INR 1000)")library(reactable)library(htmltools)library(fontawesome)htmltools::browsable(tagList( tags$button(tagList(fontawesome::fa("download"), "Download as CSV"),onclick ="Reactable.downloadDataCSV('APY_plus_adoption_indic_sf', 'APY_plus_adoption_indic_sf.csv')" ),reactable( APY_plus_adoption_indic_sf,searchable =TRUE,defaultPageSize =38,elementId ="APY_plus_adoption_indic_sf" ) ))# Yield changesYield_changes <-subset(Yield_changes, !(Yield_changes$Districts =="KISHANGANJ"))library(mapview)Yield_changes_sf <-merge(India_aoi_sf, Yield_changes, by ="Districts")mapview(Yield_changes_sf, zcol ="Yield_gain_t_ha", layer.name ="Yield gain to herbicides (tons per ha)")# Using tmaplibrary(tmap)tmap_mode("plot")Yield_changes_sf_tmap <-tm_shape(Yield_changes_sf) +tm_borders(alpha =0.5, col ="white") +tm_polygons(col ="Yield_gain_t_ha", palette ="viridis", scale =2.5, title ="Yield gain to herbicides (tons per ha)")Yield_changes_sf_tmaptmap_save(Yield_changes_sf_tmap, "figures/Yield_changes_sf_tmap.png", dpi =300, width =3.88, height =3.16)library(reactable)library(htmltools)library(fontawesome)htmltools::browsable(tagList( tags$button(tagList(fontawesome::fa("download"), "Download as CSV"),onclick ="Reactable.downloadDataCSV('Yield_changes_sf', 'Yield_changes_sf.csv')" ),reactable( Yield_changes_sf,searchable =TRUE,defaultPageSize =38,elementId ="Yield_changes_sf" ) ))```# ConclusionThe notebook has demonstrated the workflow for calculating spatially disaggregated economic surplus measures that can be used to target and prioritize agricultural interventions.# ReferencesAlston, J.M., Norton, G.W., and Pardey, P.G. 1995. "Science under scarcity: Principles and practice for agricultural research evaluation and priority setting." CAB International. Wallingford, UK.
